Metabolomics fingerprinting of thyroid malignancies: a GC/MS-based approach for subtype classification and biomarker discovery

TL;DR

This study uses metabolomics to identify unique metabolic signatures in different types of thyroid cancer, which could help in diagnosis and understanding disease progression.

Contribution

The study introduces a GC/MS-based metabolomics approach to classify thyroid cancer subtypes and discover potential biomarkers.

Findings

Distinct metabolic signatures were identified for papillary, follicular, and medullary thyroid carcinomas using plasma metabolomics.

Key metabolites like linolenic acid and arachidonic acid were reduced in thyroid cancers, while glutamine and methionine were elevated in follicular and papillary types.

A Random Forest model achieved high accuracy (AUC of 0.956) in classifying cancer subtypes based on metabolic profiles.

Abstract

Thyroid cancer encompasses distinct histological subtypes, each potentially associated with unique metabolic characteristics. However, the comprehensive metabolic reprogramming underlying these malignancies remains insufficiently characterized. Hence, this study aimed to identify untargeted metabolomics alterations in plasma samples from patients diagnosed with papillary thyroid carcinoma (PTC), follicular thyroid carcinoma (FTC), medullary thyroid carcinoma (MTC), and healthy controls, to elucidate potential metabolic signatures associated with each cancer type.

Plasma samples from patients with PTC (n = 14), FTC (n = 8), and MTC (n = 15), along with samples from healthy subjects (n = 15), were collected for untargeted metabolomics analysis using gas chromatography-mass spectrometry (GC/MS). Multivariate and univariate analyses were performed for diagnostic assessment using MetaboAnalyst, SIMCA software, and R packages.

A total of 61 metabolites were annotated across all plasma samples. Multivariate analyses, including partial least squares discriminant analysis (PLS-DA) and orthogonal PLS-DA (OPLS-DA), revealed distinct group separations and demonstrated robust model performance. One-way ANOVA followed by Tukey’s HSD and variable importance in projection (VIP ≥ 1) highlighted 35 significantly altered metabolites. Among these, linolenic acid (q = 4.76E-13) and arachidonic acid (q = 1.39E-12) showed substantial reductions across all thyroid cancer subtypes. Conversely, glutamine (q = 1.14E-10), methionine (q = 2.54E-09), and 2-hydroxybutanoic acid (q = 1.49E-07) were elevated in FTC and PTC. A Random Forest analysis further highlighted linolenic, linoleic, arachidonic acids, methionine, glutamine, and pyruvic acid, as crucial discriminative elements, achieving a macro-averaged AUC of 0.956 in multi-class classification.

This plasma metabolomics study reveals distinctive metabolic signatures associated with different thyroid cancer subtypes, suggesting potential biomarkers for differential diagnosis. These findings underscore the importance of metabolomics in enhancing subtype differentiation and provide insight into metabolic pathways associated with disease progression.

The online version contains supplementary material available at 10.1186/s12885-025-15073-0.

Figures